This great sociology blog I’ve just started reading, Org Theory, has a post on The Future of Game Theory. Good article, 3 things to be optimistic about, 3 problems it still has. I’d like to add two points as I see it after the break.
From the article:
What game theory still has going for it:
1. It is a terrific way to describe the structure of many multi-person decision making scenarios, especially certain types of market competition with fixed technologies and firms, collective action, and more. This is true whether or not one applies equilibrium concepts. (I heard Schelling say that the best thing about game theory was the invention of the extremely useful game matrix.)
What hurdles still exist:
4. People just don’t conform to the axioms of rational choice. This is not a death blow to game theory because classical game theory will always provide important an normative benchmark. However, until the models more accurately reflect actual decision making, game theory as descriptive analysis will always be constrained.
This is going to go with my second comment, but I think that modeling individual agents deeply embedded in social structures using Game Theory is a no-go. However, as with the first point, I think it is fantastic for modeling the behavior of non-human entities that tend to (or are designed to) optimize along a single dimension. Though there isn’t a lot of cross-communication, biologists are doing amazing things using Game Theory equilibrium to talk about evolution. And it is no accident – the scorched earth relationships exhibited in the theory for something like the Ultimatum game doesn’t resemble day-to-day human interactions between strangers, but certainly show up as the relationship between a wolf and a deer, or a deer and a plant. The same uses of Game Theory can be said of computer networks, where a lot of interesting work is going on in figuring out how to get computer networks to share resources conditional on other computer network’s choices – this is especially useful since the decision-making can (must, in fact), be pre-supposed and programmed. This pre-supposition for people is a problem we’ll come back to in a second.
And of course, there is that ultimate, one-dimensional entity, the profit-maximizing corporation. I’m just now getting into org theory, and I know they already have a thing about the relationship between Austrian Economics, Institutional Analysis and Game Theory (some things I know nothing about), but all I can say is that the models are fantastic ways of viewing interactions between the modern-day rationality calculator of the LLC. Even papers that don’t explicitly use the modeling call upon the ideas implicity, such as this classic about bankruptcy costs.
[4...]Some changes could be quite simple to make, e.g., incorporating Prospect Theory or quasi-hyperbolic discounting and then doing standard equilibrium analysis. Some changes would be more ambitious, e.g., drawing upon research in neuroscience to better capture new insights into how the mind works when making decisions.
I’m highly skeptical of this. (Put aside the idea of of Game Theory being a normative benchmark for people, which is chilling.) I’m skeptical if only because of the Underground Man, who phrased the futility with a sneering research summary back in the 1860s:
Consequently we have only to discover these laws of nature, and man will no longer have to answer for his actions and life will become exceedingly easy for him. All human actions will then, of course, be tabulated according to these laws, mathematically, like tables of logarithms up to 108,000, and entered in an index; or, better still, there would be published certain edifying works of the nature of encyclopaedic lexicons, in which everything will be so clearly calculated and explained that there will be no more incidents or adventures in the world.
Then — this is all what you say — new economic relations will be established, all ready-made and worked out with mathematical exactitude, so that every possible question will vanish in the twinkling of an eye, simply because every possible answer to it will be provided. Then the “Palace of Crystal” will be built.
(You can imagine using tables and vectors of logarithms for hyperbolic discounting, and, bonus for the nerds, the Crystal Palace utopia that so freaked out Dostoevsky was destroyed to halt the Nazis. So much for progress.)
But let’s look at it: The problems with applying the Game Theory to human behavior isn’t that the math isn’t tight enough, or that some curves should be concave when they are currently convex, etc. It’s that Game Theory requires people to take a predetermined, exogenous utility function that is independent of the actual act of acting and draw their choices from it. This is problematic, at best. The defunct blog antigram put it best as a first approximation:
There are really two points to be made here. Firstly, game theory cannot think a situation in which each individual is involved in a multiple number of games at the same time. For instance, it might be the case that my fellow prisoner [in the Prisoner;s Dilemma] is rumored to be tight with the Syndicate, as it might also be the case that he is engaged to marry my sister. If either of these is true, then a serious negative utility might be supposed to reside in me betraying him. But how serious a negative utility, and to what degree of magnitude? How tight really is he? And do I really want my sister marrying a common criminal? These are questions that I would have to answer on the basis of something other than purely instrumental reason, since the outcome resulting from any possible choice I could make here depends on a number of unaccountable uncertainties. Ultimately, I am forced to judge for myself as to what the best course of action would be, not just irrespective of what other’s are choosing, but also irrespective of any supposedly already decided utility. This is the basic point of Kant’s justly-famed essay What is Enlightenment, and it is the basic point that game theory denies.
We won’t start hacking into the problems of instrumental reason as a basis for a modeled human life, but I’d like to add that it’s not a accident that Adorno and others all thought of instrumental reason as synonymous with domination – not as a means of achieving free-floating utility but as a means of control over nature (and a form of irrationality at that). This embeds it in a way that simply isn’t just conditional on others reactions, and one that isn’t going to be solved by freshman sitting under a CAT scan and trading coins.